iFeeling Daily
Daily curated AI insights you can't miss.
MolmoMotion: Language-guided 3D motion forecasting from video
MolmoMotion predicts future 3D point trajectories from a single video frame, language instruction, and query points, using a class-agnostic, view-stable representation. It outperforms existing forecasting methods on a new benchmark, accelerates robot policy learning (76.3% vs 56% pick-and-place success), and improves video generation motion quality. The model, 1.16M-video dataset, and benchmark are openly released.
How UK SMBs Are Building with Google Cloud AI
UK midsize businesses are adopting AI through Google Workspace and Cloud tools, not custom models – 71% report saving time on routine tasks. For builders, the key insight is that embedding AI into existing productivity workflows and offering free upskilling programs is driving adoption faster than pure model capability.
The Local Coding Stack: 5x Speedup for Free
The local coding stack has matured to the point where models like Qwen 3.6 35B-A3B and Pi agent harness are genuinely viable for daily coding work, trading some capability for privacy, zero cost, and offline operation.
The Golden Age of AI Applications
The article argues that the golden age of AI applications hinges on mastering three disciplines—model selection, loop design, and evaluation—rather than just model performance, with regulatory and market signals from Fable, Nadella, and Salesforce reinforcing the shift.
A CEO’s Cost of Capital Advantage
A short but sharp argument that a founder's personal cost of capital is the hidden variable driving both how much they raise and how much they keep, using Elon Musk's trajectory as the extreme example of the flywheel in motion.
OpenAI Supports EU Code of Practice on AI-Generated Content Transparency
This article explains why OpenAI supports the EU's new Code of Practice on AI-generated content transparency, detailing their multi-layered provenance approach using C2PA metadata and SynthID watermarks. It's a concise look at the practical challenges and ecosystem-wide cooperation needed to build trustworthy AI content.
Biggest Mysteries in Physics: Antimatter, Dark Energy & ToE – Don Lincoln
Don Lincoln, a particle physicist at Fermilab, takes you on a deep dive into the universe's greatest unsolved puzzles—antimatter, dark energy, and the quest for a theory of everything. A clear, grounded conversation that connects experimental physics with the most profound questions about reality.
Building Grok Imagine: Video Models from Scratch at xAI
Ethan He, former NVIDIA Cosmos engineer who joined xAI to build Grok Imagine from scratch, explains why data bugs matter more than model architecture, how audio-video alignment is harder than text-video, and why language models will unlock the next wave of video generation. A rare look at the real engineering under a frontier multimodal model.
Cohere Releases North Mini Code: 30B MoE Model for Agentic Coding
Cohere's North Mini Code is a 30B MoE model optimized for agentic coding, featuring a novel two-stage SFT and asynchronous RLVR pipeline. It achieves 33.4 on the Artificial Analysis Coding Index, outperforming models like Qwen3.5 and Gemma 4. The technical report details cross-harness training that yields free improvements (e.g., 61.0% pass@1 on SWE-Bench with mini-SWE-agent) and online RL that cuts invalid tool calls by rewarding only valid outputs.